Family history of cancer is an important possible indicator of inherited cancer susceptibility, which has helped identify individuals and families at high risk of inherited cancers in research studies and clinical practice. While there are also various potential uses of family history of cancer data in cancer surveillance, the completeness and accuracy of family history of cancer data collected from the general population is unclear. In an effort to evaluate the feasibility of conducting a national surveillance study to determine the prevalence of family history of cancer in the U.S. population, the Risk Factor Monitoring and Methods Branch will undertake a pilot study, entitled the Family Health Study, that examines issues of data quality. In this study, a family history of cancer questionnaire (FHCQ) will be developed for surveillance purposes and administered to a random digit dial (RDD) sample of households in the state of Connecticut. Positive and negative reports of common cancers in the respondent's families will be validated against records of the Connecticut Tumor Registry (CTR) and other data sources.
The objectives are to: 1) assess the agreement between respondent reports of specific cancers in first and second degree relatives and medical record-based reports, as measured by percent concordance; 2) quantify the sensitivity, specificity and predictive value of the FHCQ by cancer site; 3) evaluate the possible predictors of reporting accuracy, including cancer site, year of diagnosis, kinship relation of the relative to the respondent and the frequency and quality of their contact, overall family cohesiveness, respondent's own history of cancer, and demographic factors; 4) describe the completeness and reliability of family structure data.
Validation of selected relatives' cancer status will be done through data linkage to the Connecticut Tumor Registry, other selected cancer registries, the National Death Index, Medicare claims data bases, state death certificate registries, or by obtaining consent to review available medical records from physicians and health care facilities. Self-reports of respondents' cancer status will also be validated since this may be a predictor of ability to accurately report family history. A pre-established tracing algorithm will be used to triage cancer reports into the medical records systems where true cancer status is most likely to be verified by the highest quality data. Validated cancer outcomes will be assigned ICD-9 codes by a nosology team following a double-blinded protocol. A certainty level will be assigned to each cancer outcome based on the type of confirming medical record, with evidence of microscopic confirmation of malignancy considered the most certain. Statistical analyses to determine FHCQ1 sensitivity, specificity, and predictive value, will be performed, accounting for level of certainty. Predictors of cancer reporting accuracy will be examined using multivariate regression.
